Light Metals 2008
WIRELESS MEASUREMENT OF DUCT TEMPERATURES ON ALUMINUM SMELTING POTS:
CORRELATION TO ROOFLINE HF CONCENTRATION
Daniel Steingart1,2, James W. Evans2,1, Andrew Redfern3, Paul K. Wright3,1, Neal Dando4, Weizong Xu4, Michael Gershenzon4, Henk Van
der Meyden5
1
Wireless Industrial Technologies, 4096 Piedmont Ave #193, Oakland, CA 94611
2
Dept. of Materials Science and Engineering, Univ. CA, Berkeley, CA 94720
3
Dept. of Mechanical Engineering, Univ. CA, Berkeley, CA 94720
4
Alcoa Technical Center, 100 Technical Drive, Alcoa Center, PA 15069-0001
5
Alcoa Mount Holly, P.O. Box 1000, Goose Creek, SC 29445
Keywords: Hall-Héroult cells, Fluoride emissions, Temperature sensing, Wireless sensing
Abstract
Previous papers described wireless devices (motes) for
measuring the temperature of the gas in ventilation ducts from
smelting pots, and the heat flux through the pot shells. This paper
describes more extensive, longer term measurements of duct
temperature carried out on ten pots at a smelter in 2006.
Measurements were carried out for 43 days and 1.3 million
temperatures were recorded, along with simultaneous real-time
measurements of roofline HF concentration. The duct motes were
self-powered using thermoelectric generators operating off
temperature differences between the duct gasses and the
surrounding air. Temperatures measured by the motes reflected all
pot-work such as removal of cover panels or opening of end
doors. Even after elimination of temperatures during active pot
work, significant variations in duct temperature were observed,
both from pot-to-pot and from day-to-day on individual pots. The
variations in pot duct temperature exhibited a good correlation
with roofline HF concentrations.
Introduction
Previous papers1,2 have described preliminary work on the
monitoring of duct gas temperatures and shell heat fluxes on
operating pots at Eastalco. While the results were encouraging, for
example in showing that motes were unaffected by the strong
magnetic field, these tests were too short and on too few pots to
establish the technical feasibility of wireless monitoring.
This report presents an extended term in-plant
demonstration of self-powered self-networking wireless process
measurement technology in operating aluminum smelters.
Wireless duct temperature measurements were carried out on ten
pots at an Alcoa plant in the Fall of 2006 for a period of 43 days.
Wireless Devices Used
The principal wireless device used is illustrated in Fig. 1. It
consists of a thermistor projecting from the end of a copper rod,
both being positioned within the duct through a hole in the wall of
the duct. The rest of the device remained outside the duct and
consists of a thermoelectric generator (TEG), a heat sink and the
wireless “mote” (plus other electronics contained within the black
box towards the top of the photograph). The copper rod served to
conduct heat to the TEG which was positioned between the
aluminum plate at the base of the copper rod and the heat sink, the
latter serving to establish a temperature differential across the
TEG. The TEG provided enough power to run both the mote‟s
transceiver and the sensor electronics converting the output from
the thermistor to an input signal for the mote which then
wirelessly relayed the thermistor reading to a laptop computer set
to collect the data. The laptop then relayed the data to Wireless
Industrial Technologies (WIT) in California via a Sprint cell
„phone card. The devices were assembled by WIT using
commercial and custom components.
Additional wireless motes were placed in the proximity
of the ten pots under study to provide a robust network that would
not suffer interference from tapping ladles etc. Some of these
“repeater” motes were also powered by TEGs (working off the hot
surfaces of the cell end wall); others were powered by batteries.
Fig. 2 illustrates the placement of the motes in the
potroom and is a schematic overhead view showing the ten pots
under study with ten devices mounted on their ducts for gas
temperature sensing. Also shown are the locations of the repeater
motes. The magenta square represents a “base station”, wired to
the laptop in a shed adjacent to the potroom, that received the
wireless data from the ten temperature sensing motes, either
directly or via other motes.
Results
344 pot-days of data were collected with each duct temperature
reported approximately every 25 seconds for a total of 1.3 million
temperature measurements. Additional data concerning mote
connectivity and voltage were also transmitted to WIT. Minor
difficulties were encountered (such as the laptop attempting an
automatic upgrade and waiting for ten hours to be instructed by a
human to go ahead) but no insuperable problems occurred.
Fig. 3 shows the duct temperature trace for one pot for
the whole of day 13. One objective of duct temperature
measurements is detection of panel covers that are not in place.3
During this day one of the authors (HVM) deliberately removed
three cover panels and the drop in the duct gas temperature,
caused by an excess inflow of cold air, is clearly seen in the
temperature trace. Similar drops in temperature due to opening an
end door were also readily seen in the data (not shown here).
Box containing TSquare board and
Crossbow Mica2
Copper “proboscis” conducting
heat from duct gasses to the
aluminum plate
Thermistor
Sealing pad between duct
exterior and the device
Beneath this aluminum plate is
the TEG with one side in contact
with the hot plate
Finned heat “sink”
in contact with one
side of TEG
Adjustable
mounting feet
Penny
Figure 1. Wireless mote for measurement of duct temperature
Figure 2. Placement of motes on ten cells in potroom.
Figure 3. Temperature trace for one pot on a day when three cover panels were removed to see the effect on duct temperature.
Table I. Duct temperatures (0C) during “quiet” periods.
Pot #
177
175
171
176
172
173
168
174
169
170
Mean
142
135
135
134
127
125
123
121
121
121
Furthermore it was possible to discern from the
temperature trace the periods of abnormal duct temperature due
to operators working the pot, e.g by upward excursions of
temperature during an anode change as the gasses under the
hood are exposed to hot bath. One difficulty in analyzing the
temperature data to discern pot-to-pot variations was due to the
stochastic nature of such pot working. An algorithm was written
to process the data so as to tease out only the temperature data
for “quiet” periods when the cell was undisturbed (47 to 68% of
the time). The minimum, maximum and mean temperatures
(over the 43 day period) for the ten pots during quiet periods are
displayed in Table I. For each pot there were substantial
variations in duct temperature with time, over a range of about
300C probably due to imperfect replacement of cover panels or
variations in ore cover of the bath. Just as remarkable was the
variation from pot to pot in the mean temperature; the hottest pot
Max
157
149
152
147
142
137
135
137
134
135
Min
129
120
114
119
103
112
108
106
104
97.9
average gas temperature was 210C higher than the coolest. If it
can be assumed that the duct gas flow rates are nearly the same
then this would imply that the pots at the top of the table are
loosing energy, and therefore consuming electrical energy, at a
rate higher than the coolest pots. That difference amounts to a
significant fraction of a kWh/kg aluminum. A more likely
explanation is that this assumption is incorrect and the gas flow
rates of the pots towards the top of the table are lower than those
towards the bottom.
The TEGs powered their motes reliably over the entire
43+ days of the test while the batteries showed the expected
decline to 2V at which point the battery powered motes shut
down. The mote network was sufficiently robust that the loss of
the battery powered motes was insignificant.
BOREAL HF MEASUREMENTS
6
5
4
3
2
1
0
10/3/2006 12:00 10/4/2006 0:00 10/4/2006 12:00 10/5/2006 0:00 10/5/2006 12:00 10/6/2006 0:00 10/6/2006 12:00 10/7/2006 0:00 10/7/2006 12:00
Figure 4. Comparison of roofline HF measurements (upper plot) with index of duct temperature drop for the same three day
period in October, 2006.
Correlation of Duct Temperature Measurements with
Roofline HF Measurements
Another significance of the duct temperature is that it can signal
excess HF emissions.4 Gaps between, or removal of cover
panels promotes the flow of moist air into the pots. This results
in a characteristic drop in duct temperature. Moist air can react
with bath to produce HF and the absence of cover panels (or
presence of gaps) allows the escape of some of the HF into the
potroom. The wireless duct temperature data were processed at
WIT with another algorithm to detect abnormally low
temperatures and a running index of such abnormalities
calculated. Fig. 4 is a comparison between WIT‟s temperature
abnormality index and the roofline HF measurements made
above the ten pots by Alcoa personnel. Clearly duct temperature
measurement could be a tool for signaling and minimizing
fugitive HF.
Concluding Remarks
Wireless measurement of pot parameters can be a way of
improving pot performance without the hazards/costs of
stringing instrument wires around the potroom. The
investigation described here has established the technical
feasibility of such applications of wireless technology and may
lead to a reduction of the environmental impact of aluminum
smelters.
Acknowledgements
This project was funded by the Alcoa Technical Center. The
authors would also like to thank Alcoa‟s Sustainable Production
Technology and Primary Metals Environmental Management
teams for encouraging the development and publication of
environmental technology advances.
1
M. Schneider et al., Light Metals 2005, H. Kvande ed., TMS,
Warrendale, PA, 2005, p 407 & Light Metals 2006,
2
M. Schneider et al., Light Metals 2006, T. J. Galloway ed.,
TMS, Warrendale, PA, 2006, p331
3
N. R. Dando and R. Tang, Light Metals 2006, T. J. Galloway
ed., TMS, Warrendale, PA, 2006, p203
4
Neal Dando, Light Metals 2004, A. Taberaux ed., TMS,
Warrandale, PA, 2004, p245.